High-Performance Processing in a Virtualization Environment

2. The computer implemented method of claim 1, further comprising, selecting, the first appliance virtual machine based on one or more of (a) previous use of one or more appliance virtual machines, (b) load balancing between available appliance virtual machines, (c) latency in response from the available appliance virtual machines, or (d) co-location within a same host device as the client virtual machine.

3. The computer implemented method of claim 1, further comprising, upon receiving, by the first appliance virtual machine the request from the client virtual machine, load balancing the request between one or more schedulers based on use of the GPGPU or the co-processor by the client virtual machine and other client virtual machines.

4. The computer implemented method of claim 1, further comprising, prior to the transmitting, determining, in response to the system call, that an established connection between the client virtual machine and the first appliance virtual machine has switched from a first protocol to a second protocol, the first and second protocols including a high-performance transmission protocol and a low-performance transmission protocol.

6. The computer implemented method of claim 4, wherein the established connection switches to the second protocol in response to the client virtual machine migrating to the first host device from a second host device.

9. The method of claim 1, wherein the parallel computing framework library comprises the OpenCL framework library.

11. The non-transitory computer-readable medium of claim 10, wherein the operations further comprise, selecting, the first appliance virtual machine based on one or more of (a) previous use of one or more appliance virtual machines, (b) load balancing between available appliance virtual machines, (c) latency in response from the available appliance virtual machines, or (d) co-location within a same host device as the client virtual machine.

12. The non-transitory computer-readable medium of claim 10, wherein the operations further comprise, upon receiving, by the first appliance virtual machine the request from the client virtual machine, load balancing the request between one or more schedulers based on use of the GPGPU or the co-processor by the client virtual machine and other client virtual machines.

13. The non-transitory computer-readable medium of claim 10, wherein the operations further comprise, prior to the transmitting, determining, in response to the system call, that an established connection between the client virtual machine and the first appliance virtual machine has switched from a first protocol to a second protocol, the first and second protocols including a high-performance transmission protocol and a low-performance transmission protocol.

15. The non-transitory computer-readable medium of claim 13, wherein the established connection switches to the second protocol in response to the client virtual machine migrating to the first host device from a second host device.

18. The apparatus of claim 17, wherein to process the system call further comprises to, select, the first appliance virtual machine based on one or more of (a) previous use of one or more appliance virtual machines, (b) load balancing between available appliance virtual machines, (c) latency in response from the available appliance virtual machines, or (d) co-location within a same host device as the client virtual machine.

19. The apparatus of claim 17, wherein to process the system call further comprises to, upon receiving, by the first appliance virtual machine the request from the client virtual machine, load balance the request between one or more schedulers based on use of the GPGPU or the co- processor by the client virtual machine and other client virtual machines.